Control Systems m i l e r p r a in r e v y n is o Bode plots L. Lanari Monday, November 3, 2014 Outline • • • • • Bode’s canonical form for the frequency response Magnitude and phase in the complex plane The decibels (dB) Logarithmic scale for the abscissa Bode’s plots for the different contributions Lanari: CS - Bode plots Monday, November 3, 2014 2 Frequency response The steady-state response of an asymptotically stable system P(s) to a sinusoidal input u(t) = sin ω̄t is given by yss (t) = |P (j ω̄)| sin(ω̄t + ∠P (j ω̄)) P(j!) is the restriction of the transfer function to the imaginary axis P(j!)|s = j! Frequency response P (jω) � |P (jω)| ∠P (jω) gain curve (or magnitude) phase curve for ω ∈ R+ = [0, +∞) Lanari: CS - Bode plots Monday, November 3, 2014 3 Bode canonical form pole/zero representation of the transfer function with m such that • • • � � 2 2 (s − z ) (s + 2ζ ω s + ω ) 1 k � n� � n� k � � 2 F (s) = K m � 2 ) s (s − p ) (s + 2ζ ω s + ω i z nz nz i z m = 0 if no pole or zero in s = 0 m < 0 if m zeros in s = 0 m > 0 if m poles in s = 0 remarks • • • numerator and denominator are by hypothesis coprime denominator is monic K’ is not the system gain Lanari: CS - Bode plots Monday, November 3, 2014 4 Bode canonical form • the terms (s - zk) and (s - pi) are relative to ‣ ‣ • real zeros (in s = zk) real poles (in s = pi) 2 2 ) and (s2 + 2ζz ωnz s + ωnz ) are relative to the terms (s2 + 2ζ� ωn� s + ωn� ‣ complex conjugate zeros (in s = α� ± jβ� ) ‣ complex conjugate poles (in s = αz ± jβz ) with ‣ ‣ natural frequency ωn∗ = � α∗2 + β∗2 damping coefficient ζ∗ = −α∗ /ωn∗ Lanari: CS - Bode plots Monday, November 3, 2014 � = −α∗ / α∗2 + β∗2 5 Bode canonical form factoring out the constant terms s − zk = −zk (1 − 1/zk s) = −zk (1 + τk s) s − pi = −pi (1 − 1/pi s) = −pi (1 + τi s) with with τk = −1/zk τi = −1/pi with ¿i and ¿k being time constants � � 2 � � 2 2 (−z ) (ω ) (1 + τ s) (1 + 2ζ /ω s + s /ω ) 1 k k � n� � n� n� k � k � � � � F (s) = K m � 2 2 2 s i (−pi ) z (ωnz ) i (1 + τi s) z (1 + 2ζz /ωnz s + s /ωnz ) defining � � 2 � �k (−zk )� � (ωn� ) K=K 2 ) (−p ) (ω i nz i z K = [sm F (s)]s=0 for any m�0 how to compute K � � 2 (1 + τk s) � (1 + 2ζ� /ωn� s + s2 /ωn� ) 1 k � � F (s) = K m 2 /ω 2 ) s (1 + τ s) (1 + 2ζ /ω s + s i z nz nz i z Bode canonical form Lanari: CS - Bode plots Monday, November 3, 2014 6 Gains � � 2 (1 + τk s) � (1 + 2ζ� /ωn� s + s2 /ωn� ) 1 k � � F (s) = K m 2 /ω 2 ) s (1 + τ s) (1 + 2ζ /ω s + s i z nz nz i z generalized gain K = [sm F (s)]s=0 for any = 1 for s = 0 m�0 Note that • for a system with no poles (i.e. m negative or zero) we have defined as dc-gain (or static gain) � � Ks = F (s)� = F (0) s=0 if m < 0 (zeros in s = 0) we have F(0) = 0 • static and generalized gain coincide only when m = 0 K = Ks • ⇔ m=0 for an asymptotically stable system, the step response tends to the static gain Ks = F(0) Lanari: CS - Bode plots Monday, November 3, 2014 7 Bode canonical form Examples F (s) = K = F (s) = K = F (s) = K = Lanari: CS - Bode plots Monday, November 3, 2014 s−1 s−1 1 1−s = =− 2 2s + 6s + 4 2(s + 1)(s + 2) 2 (1 + s)(1 + s/2) 1 Ks = − 2 s(s − 1) 1 s(1 − s) = =− 2 2(s + 1) (s + 2) 2 (1 + s)2 (1 + s/2) 1 − Ks = 0 2 s−1 1 1−s =− 2s(s + 1)(s + 2) 2 s(1 + s)(1 + s/2) 1 − � Ks 2 8 Bode canonical form frequency response 1 F (jω) = K (jω)m � � 2 2 (1 + jωτ ) (1 + 2ζ jω/ω + (jω) /ω ) k � n� n� k � � � 2 2 i (1 + jωτi ) z (1 + 2ζz jω/ωnz + (jω) /ωnz ) has 4 elementary factors 1. constant K (generalized gain) 2. monomial j! (zero or pole in s = 0) 3. binomial 1 + j!¿ (non-zero real zero or pole) 4. trinomial 1 + 2³(j!)/!n + (j!)2/!n2 (complex conjugate pairs of zeros or poles) so first check which kind of root you have and then factor it out Lanari: CS - Bode plots Monday, November 3, 2014 9 Bode diagrams for any real value of the angular frequency ! the frequency response F(j!) is a complex number |F (jω)| magnitude of the frequency response as a function of the angular frequency ! ∠F (jω) angle or phase of the frequency response as a function of the angular frequency ! F (jω) = Re[F (jω)] + jIm[F (jω)] F (jω) = |F (jω)|e j∠F (jω) F(j!) Im[F(j!)] |F )| ω (j ∠F (jω) Re[F(j!)] |F (jω)| = � Re[F (jω)]2 + Im[F (jω)]2 Lanari: CS - Bode plots Monday, November 3, 2014 ∠F (jω) = atan2(Im[F (jω)], Re[F (jω)]) 10 Phase Phase[F.G] = Phase[F ] + Phase[G] the phase of a product is the sum of the phases and therefore � � F = Phase[F ] − Phase[G] Phase G the phase of a ratio is the difference of the phases since Phase �1� G = −Phase[G] very useful since we can find the contribution to the phase of each term and then just do an algebraic sum Lanari: CS - Bode plots Monday, November 3, 2014 11 Phase F(j!) Im[F(j!)] |F )| ω (j principal argument takes on values in (- ¼, ¼] and is implemented by the function with two arguments ∠F (jω) Re[F(j!)] for P = ® + j¯ atan2(β, α) = Lanari: CS - Bode plots Monday, November 3, 2014 � � β arctan α � � β arctan α +π � � β arctan α −π π 2 sign(β) undefined atan2 if α>0 (I & IV quadrant) if β ≥ 0 and α < 0 (II quadrant) if β < 0 and α < 0 (III quadrant) if α = 0 and β �= 0 if α = 0 and β = 0 II I III IV 12 Magnitude in order to have the same property we need to go through some logarithmic function |F (jω)|dB = 20 log10 |F (jω)| same nice properties |0.1|dB = −20 dB |10|dB = 20 dB Lanari: CS - Bode plots Monday, November 3, 2014 |F.G|dB = |F |dB + |G|dB � � �1� � � = − |F |dB �F � dB as phase |1|dB = 0 dB decibels (dB) �√ � � 2� dB ≈ 3 dB |100|dB = 40 dB |F |dB � +∞ if |F | � ∞ |F |dB � −∞ if |F | � 0 13 Logarithmic scale we use a logarithmic (log10) scale for the abscissa (angular frequency !) log10(!) becomes a straight line if ! is in a logarithmic scale 20 log10(|x|) a decade corresponds to multiplication by 10 20 0 −20 −40 −60 −3 10 −2 −1 10 0 10 x in logarithmic scale 1 10 10 very useful when we add |x| 10 different contributions 5 20 log10(|x|) 0 −3 10 Monday, November 3, 2014 −1 0 10 x in logarithmic scale 1 10 10 20 0 −20 −40 −60 Lanari: CS - Bode plots −2 10 1 2 3 4 5 6 x in linear scale 7 8 9 10 14 Logarithmic scale advantages • • quantities can vary in large range (both ! and magnitude) easy to build the magnitude plot in dB of a frequency response given in its Bode canonical • form from the magnitudes of the single terms easy to represent series of systems |F(jω)| 300 200 100 0 2 4 6 8 10 linear scale same data 60 |F(jω)|dB different scales for abscissa and 0 40 20 0 −20 ordinates 1 2 3 4 5 6 linear scale 7 8 9 10 |F(jω)|dB 60 40 20 0 −20 −1 10 Lanari: CS - Bode plots Monday, November 3, 2014 0 10 logarithmic scale 1 10 15 Bode diagrams |F (jω)|dB ∠F (jω) magnitude in dB of the frequency response as a function of the angular frequency ! with logarithmic scale for ! angle or phase of the frequency response as a function of the angular frequency ! with logarithmic scale for ! we need to find the magnitude (in dB) and phase for the 4 elementary factors 1. constant K (generalized gain) 2. monomial j! (zero or pole in s = 0) 3. binomial 1 + j!¿ (non-zero real zero or pole) 4. trinomial 1 + 2³(j!)/!n + (j!)2/!n2 (complex conjugate pairs of zeros or poles) for ω ∈ R+ = [0, +∞) Lanari: CS - Bode plots Monday, November 3, 2014 16 Constant Im √ K K3 = − 10 K1 = √ 10 Re √ | 10|dB = 10 dB √ | − 10|dB = 10 dB |0.5|dB ≈ −6 dB magnitude 20 log10 |K| Magnitude (dB) K2 = 0.5 20 |K 1 | dB , |K 3 | dB 10 0 |K 2 | dB −6 −10 −2 10 √ ∠0.5 = 0◦ 0 10 frequency (rad/s) 1 10 2 10 0 K1, ! K2 ! phase ∠K Phase (deg) ∠ 10 = 0◦ √ ∠ − 10 = −180◦ = −π −1 10 −50 −100 −150 K3 ! −180 −2 10 Lanari: CS - Bode plots Monday, November 3, 2014 −1 10 0 10 frequency (rad/s) 10 1 10 2 17 Monomial - Numerator Im j! j! 90° |jω|dB = 20 log10 ω Re log scale |jω|dB = 20x magnitude Magnitude (dB) 40 20 dB/dec 20 0 −20 phase Phase (deg) −40 −2 10 −1 10 0 10 frequency (rad/s) 1 10 2 10 90 80 60 40 20 0 −2 10 Lanari: CS - Bode plots Monday, November 3, 2014 −1 10 0 10 frequency (rad/s) 10 1 2 10 18 Monomial - Denominator from properties of log and phase magnitude Magnitude (dB) 40 20 0 -20 dB/dec −20 −40 −2 10 −1 10 0 10 frequency (rad/s) 1 10 2 10 phase Phase (deg) 0 −20 −40 −60 −80 −90 −2 10 Lanari: CS - Bode plots Monday, November 3, 2014 −1 10 0 10 frequency (rad/s) 10 1 2 10 19 Binomial - Numerator magnitude 1 + j!¿ |1 + jωτ |dB = 20 log10 � 1 + ω2 τ 2 approximation wrt the cutoff frequency 1/|τ | (corner frequency) 1 if ω � 1/|τ | � 1 + ω2 τ 2 ≈ √ 2 2 ω τ if ω � 1/|τ | and therefore |1 + jωτ |dB at the cutoff frequency ∗ ≈ ω = 1/|τ | 0 dB if ω � 1/|τ | 20 log10 ω + 20 log10 |τ | if ω � 1/|τ | |1 + jτ /|τ | |dB = 20 log10 √ 2 ≈ 3 dB two half-lines approximation: 0 dB until the cutoff frequency, + 20dB/decade after Lanari: CS - Bode plots Monday, November 3, 2014 20 Binomial - Numerator 1 + j!¿ phase depends on the sign of ¿ Im ¿<0 j!¿ Im Re 1 Re 1 ¿>0 j!¿ see how the phase changes as ! increases Lanari: CS - Bode plots Monday, November 3, 2014 21 Binomial - Numerator case ¿ > 0 case ¿ < 0 ∠(1 + jωτ ) ∠(1 + jωτ ) ≈ ≈ 0 phase depends on the sign of ¿ 1 + j!¿ ω � 1/|τ | if π 2 ω � 1/|τ | if 0 if − π2 if and τ >0 ω � 1/|τ | ω � 1/|τ | and τ <0 the two asymptotes are connected by a segment starting a decade before (0.1/| ¿ | ) the cutoff frequency and ending a decade after (10/| ¿ |). The approximation is a broken line. ¼/2 1/| ¿ | 0 -¼/2 at the cutoff frequency Lanari: CS - Bode plots Monday, November 3, 2014 ω ∗ = 1/|τ | ∠(1 + jτ /|τ |) = π 4 if τ >0 − π4 if τ <0 22 Binomial - numerator 1/| ¿ | 1 + j!¿ magnitude Magnitude (dB) 40 30 20 10 3 0 −10 frequency (rad/s) 0.1/| ¿ | 10/| ¿ | ¿>0 phase Phase (deg) 90 45 0 frequency (rad/s) 0.1/| ¿ | 1/| ¿ | 10/| ¿ | ¿<0 phase Phase (deg) 0 −45 −90 frequency (rad/s) Lanari: CS - Bode plots Monday, November 3, 2014 23 Binomial - denominator 1/| ¿ | 1 /(1 + j!¿) magnitude Magnitude (dB) 10 0 −3 −10 −20 −30 −40 frequency (rad/s) 10/| ¿ | 0.1/| ¿ | ¿>0 phase Phase (deg) 0 −45 −90 frequency (rad/s) 0.1/| ¿ | 1/| ¿ | 10/| ¿ | ¿<0 phase Phase (deg) 90 45 0 frequency (rad/s) Lanari: CS - Bode plots Monday, November 3, 2014 24 Trinomial magnitude � � 2 � � �1 + 2 ζ (jω) + (jω) � � ωn ωn2 � = = � � 2 � � �1 − ω + j2ζ ω � � ωn2 ωn � �� 1− ω2 ωn2 �2 � + 4ζ 2 ω2 ωn2 � approximation wrt !n |TRINOMIAL| |TRINOMIAL|dB Lanari: CS - Bode plots Monday, November 3, 2014 ≈ ≈ 1 �� � 2 2 ω = ω2 n ω2 2 ωn if ω � ωn if ω � ωn 0 dB if ω � ωn 40 log10 ω − 20 log10 ωn2 if ω � ωn 25 Trinomial in ! = !n the magnitude | TRINOMIAL | is equal to 2 | ³ | |ζ| 0 0.5 √ 1/ 2 ≈ 0.707 |T RIN |dB in ωn −∞ 0 dB 3 dB 1 6 dB large variation of the magnitude in ! = !n depending upon the value of the damping coefficient ³ no approximation around the natural frequency !n Lanari: CS - Bode plots Monday, November 3, 2014 26 Trinomial ") How does a generic complex root "( varies in the plane as a function of ! &' " !! " ) ! " (! " # " # "# !" $# Phase � ζ (jω) ∠ 1 + 2 (jω) + ωn ωn2 2 � = 0 π −π % # ω � ωn if if ω � ωn and ζ≥0 if ω � ωn and ζ<0 transition between - ¼ and ¼ is symmetric wrt !n and becomes more abrupt as | ³ | becomes smaller. When ³ = 0 the phase has a discontinuity in !n Lanari: CS - Bode plots Monday, November 3, 2014 27 Trinomial - numerator magnitude Magnitude (dB) 60 40 0.7 20 1 0 0.5 −20 0.3 0.1 0 −40 frequency (rad/s) 0.1 !n 10 !n !n phase ζ≥0 Phase (deg) 180 0.1 135 0.3 0.7 90 1 45 0 0.5 0 frequency (rad/s) 0.1 !n !n 10 !n phase ζ<0 Phase (deg) 0 −45 0.01 1 −90 −135 −180 frequency (rad/s) Lanari: CS - Bode plots Monday, November 3, 2014 28 Trinomial - denominator magnitude Magnitude (dB) 40 0 20 0 1 −20 −40 −60 frequency (rad/s) 0.1 !n !n 10 !n phase ζ≥0 Phase (deg) 0 0 −45 1 −90 −135 −180 frequency (rad/s) 0.1 !n !n 10 !n phase ζ<0 Phase (deg) 180 135 90 45 1 0.01 0 frequency (rad/s) Lanari: CS - Bode plots Monday, November 3, 2014 29 Trinomial When | ³ | = 1 the trinomial reduces to a product of identical binomials (real roots) roots � � = 2 ζ s 1+2 s+ 2 ωn ωn −ωn if ζ=1 ωn if ζ = −1 � = ζ=±1 � s 1± ωn �2 and therefore the magnitude and phase coincides with that of a double binomial with corner frequency 1 = ωn |τ | that is in ! = !n when | ³ | = 1 2 × (3 dB) = 6 dB 2 × (−3 dB) = −6 dB (numerator) (denominator) example: MSD system with critical value for the damping (µ2 = 4km) Lanari: CS - Bode plots Monday, November 3, 2014 30 Trinomial √ if |ζ| < 1/ 2 ≈ 0.707 the magnitude of a trinomial factor at the denominator has a peak at the resonance frequency 1 � |F (jωr )| = 2|ζ| 1 − ζ 2 resonance peak � ωr = ωn 1 − 2ζ 2 Magnitude (dB) 20 0 0.5 0.3 −20 0.1 0.01 34 0.01 14 5 1 −20 −40 frequency (rad/s) anti-resonance peak Lanari: CS - Bode plots Monday, November 3, 2014 Magnitude (dB) (similarly for the anti-resonance peak) frequency (rad/s) resonance peak 31